Tip

Read this first.

Amazon product reviews: analyzing user feedback through sentiment score and issue categorization

Recently, I picked up running as a hobby. I was debating whether to buy a running vest to hold my phone and keys while outdoors, so I decided to utilize my data skills to analyze this Amazon product.

The task

• Scrape product reviews from the Amazon website (I used a mix of BeautifulSoup and Selenium to navigate all the pages).
• Create a dataframe to store all reviews and their rating scores.
• Clean and explore the data.
• Add sentiment analysis with BERT pre-trained NLP model.
• Visualize and review results.
• Compare rating score (1 to 5 stars) to BERT score (1 to 5 points) as in distribution and correlation between the two.
• Identify more nuanced reviews and uncover additional insights.
• Visualize reviews' keywords.
• Categorize issues using targeted keywords on reviews.

Findings

• Currently, we have 73 reviews for this product with an average rating of 4.46 stars.
• This running vest's strongest asset is its features, such as holders and reflective strips for extra comfort when running.
• Size seems to be the main issue highlighted by customers. This is visible in all techniques employed.
• Through a sentiment analysis of our reviews and a comparison of rating scores, we picked up more nuanced user feedback about sizing issues and comfort in warm weather.
• On a positive note, the customer support team was praised for providing high quality service.
• Through a keyword analysis, we noticed popular words used in our reviews are mostly positive, with focus on product features . We have a few negative mentions concerning sizing.
• By categorizing reviews based on issues, size is mentioned in 4% of our reviews, while there are no complaints related to price or faulty items.

Thank you for checking out this repo! 🌟